Risk Reporting Practices of Listed Companies: Cross- Country Empirical Evidence from the Auto Industry

Prior literature has shown that corporate transparency is linked to a firm’s competitiveness. The frequently noted vagueness and inadequacy of reported risk disclosures have been accompanied by calls for industry-specific studies. This paper aims to examine the disclosure informativeness of leading multinational automobile firms worldwide regarding firm-specific risks, namely company risk and company size. Applying a content analysis to examine the prevalent disclosure context in the automotive industry, we analyze and classify the annually reported risk statements of 34 multinationals quoted within the NASDAQ OMX Global Automobile Index. These corporations are headquartered in 10 countries, located in the U.S. as well as in the E.U. and in Asia. Based on an employed Wilcoxon signed rank test, our results show that automobile multinationals favor revealing fewer (more) forward-looking and bad-news risk disclosures as compared to past-looking and good-news risk statements. Our findings further indicate that bigger and riskier automobile multinationals do not reveal larger amounts of risk information. This finding is inconsistent with disclosure theory and provides insight into less complex risk reporting practices that affect investor perceptions of risk by omitting risk information. This paper also provides new empirical evidence regarding the association between company risk and corporate risk disclosing practices in the auto industry; our data show that company size did not play a significant role. With our concentration on explaining current risk reporting practices in the automotive sector, these results are robust with respect to firm-specific risk variables, thus the findings may prove usable for policy frameworks.


INTRODUCTION
The principles of risk, risk reporting, and risk management have recently attracted considerable attention as the world economy has become increasingly competitive (Lobo et al., 2019). Transparency is widely considered critical for individual firms as well as in terms of national com- petitiveness (Wyld, 2009). Corporate transparency in financial reporting refers essentially to disclosing risks in corporate annual reports (Kang & Gray, 2019). The call for transparent corporate financial reporting is widely accepted by regulators worldwide as an important theme in economic-political debates in a global context (Transparency International, 2014; Kang & Gray, 2019). This is due to the numerous recent corporate scandals which have been blamed on multinational firms hiding their risky and questionable operating and business activities. The most recent financial crisis has also raised the concern that mandatory disclosures also portray firm risk non-transparently (Badia et al., 2019). Consequently, the external pressure on companies to become more transparent toward their stakeholders and investors has increased (Manes-Rossi, 2018). Previous literature dedicated to disclosed risks in the annual reports of non-financial companies has usually been conducted within one country as a cross-industry study of companies listed within a particular stock market index. However, the ongoing call has not been heeded for broader risk disclosure studies. The studies which have been conducted have been found particularly lacking in terms of cross-country and industry-specific risk disclosure studies that could improve the overall perception of business risk reporting practices.
This paper seeks to begin to fill this knowledge gap as well as to supplement the existing risk reporting literature in other ways. First, this article adds to the relatively limited cross-country literature related to risk disclosure in the automotive industry (generally conducted only as panel studies within various industries, e.g., Campbell et al., 2014, Hope et al., 2016. Second, the explanation of the links between two firm-specific characteristics could enhance the understanding of the risk disclosing practices dedicated to the auto industry: i) a level of company size ii) the level of company risk. These two characteristics are evaluated in terms of quantity of iii) total iv) total financial v) and total non-financial disclosed risk statements in the 34 recent annual reports of the biggest and liquid automakers headquartered and listed in 10 countries, located in the U.S. as well as in the E.U. and in Asia. Third, this study explains the prevalent nature of the disclosed risk information, and fourth, sheds a degree of light on whether more i) monetary or non-monetary risk disclosures, ii) past or forward-looking risk disclosures, iii) and positive news or negative news risk disclosures are disclosed by automakers, with our results seeking to provide policy and practical implications. These findings may help accounting policymakers and academics broaden their understanding of risk reporting and educate financial statement users to consider the implications and act accordingly. The research goals of this paper are namely to i) test for a linkage between the level of company size and the volume of risk disclosure, ii) test for a relation between the level of company risk and the volume of risk disclosure, iii) examine the characteristics of the risk disclosure considering the type of risk, the proportion of good and bad news risk sentences, and time orientation. The automotive industry was chosen particularly for its large global economic footprint and for its global industry interconnectedness ( Bank, 2019). Further, the automotive sector is of current interest to policymakers since it contributes substantially to greenhouse gas emissions (Nadirov et al., 2020).
The paper is organized as follows. Section 2 presents the theoretical framework and hypotheses. The research design is described in Section 3. The research results are presented in Section 4, and Section 5 concludes the paper with a review of the empirical findings and concluding remarks.

THEORETICAL BACKGROUND AND HYPOTHESIS DEVELOPMENT
Risk disclosure is information that explains particular corporate risks and the expected exposure of firms to current and future corporate performance scrutiny (Dey et al., 2018;Miihkinen, 2012

Size
It is argued that larger firms can afford to reveal risk information because the proportionate cost will be less when compared to smaller companies (

Hypothesis H1a (H1a): There is a positive association between company size and the total number of risk disclosures.
Hypothesis H1b (H1b): There is a positive association between company size and the total number of financial risk disclosures.
Hypothesis H1c (H1c): There is a positive association between company size and the total number of non-financial risk disclosures.

Level of risk
As per Linsley & Shrives (2006) and in accordance with agency theory, it should be noted that companies with greater risk levels would prefer to close less information on risks because the directors need to clarify the reasons behind the company's higher risk. However, some of the directors are unwilling to share their riskiness (Linsley & Shrives, 2006). Thus, the association between the level of risk disclosure and company risk is considered null, as we explain next. Relying on the risk management and agency theory, we can expect that there will be a positive link between disclosing risk information and the level of company risk. The early studies offer two opposing views of the linkage between the company level of risk and level of disclosure: positive association ( Hypothesis H2a (H2a): There is no association between company risk and the total number of risk disclosures.
Hypothesis H2b (H2b): There is no association between company risk and the total number of financial risk disclosures.
Hypothesis H2c (H2c): There is no association between company risk and the total number of non-financial risk disclosures.

Economic sign (monetary/non-monetary)
The importance of monetary value (quantified estimates) of disclosed risk is essential to increase the quality of risk reporting (Linsley & Shrives, 2006). According to Frame (2003), the quantifications of risk can be challenging due to data unavailability. However, this situation depends on highly subjective or non-subjective inputs. Linsley

Type of measure (past/future)
It has been emphasized that future risk information is more beneficial in the decision-making of stakeholders than past information (Linsley & Shrives, 2006;Ntim et al., 2013;Marzouk, 2016). It is because the future is uncertain, and the wrong information can mislead the people who relied on them. The results of Ntim et al., (2013) pointed out that past-looking disclosures were more frequent than forward-looking disclosures. Therefore, the hypothesis H4 is as follows: Hypothesis H4 (H4): The number of future risk disclosures will be considerably less than the number of past risk disclosures.

Outlook sign (good/bad/neutral)
Commonly, directors do not want to share bad news with external parties (Linsley & Shrives, 2006), while good news disclosures generate lesser stock price reactions than bad news disclosures (Skinner, 1994). If managers fail to report bad news in a timely manner, reputational losses may be incurred, and stockholders may not follow firms reputed to withhold bad news (Skinner, 1994). Consequently, as depicted in attribution theory, directors can disclose less positive risk information (Marzouk, 2016 (Skinner, 1994). This indicates that it is hard to predict the amounts of bad and good risk news. Therefore, the following is hypothesized.
Hypothesis H5 (H5): The number of good and bad risk disclosures will not differ significantly from each other.

Data sample
The

Variable description
The level of company size and risk can be measured in several ways. In this study, to test the hypotheses, the level of company size and the company level of risk need to be measured. Turnover and market value have been selected to measure the level of company size in this paper. Turnover has been sourced directly from the annual report as of 31 Systematic risk (3Y monthly beta) and stock volatility (3Y monthly standard deviation) of the company were selected to measure the level of company risk. The systematic risk (3Y monthly beta) was sourced from Yahoo Finance (2019), and stock volatility was calculated using monthly company market prices, in line sourced from Yahoo Finance (2019). Beta measures systematic risk of a company and is calculated by estimating the company's market price volatility relative to the stock market index. Statistically described, beta represents a line slope by regressing data points from the returns of an individual stock against those of the market. A beta of 1 indicates that the company's stock price moves in line with the market, the beta of less than one means that the company is less risky than the stock market, and a beta of greater than one indicates that the company is riskier than the stock market. Companies with low beta do not rise and fall as much as the market, while high beta companies tend to behave opposite, which means they are more volatile. Beta in this paper is the 36 months' average of a monthly change in a company's price compared to the monthly price change of the market. Stock volatility of the company is an average of 36 months of quarterly standard deviations of company market price returns.

Methods
This risk disclosure study aims to examine the risk reporting practices of automakers to answer the calls for more risk reporting and industry-specific risk disclosure studies.
To test the main hypotheses derived in the literature review section, we conducted the following procedure. We applied content analysis that is considered to be a widely used method for analyzing disclosures in accounting research (Abid & Shaique, 2015) in this empirical study. In line with Linsley & Shrives (2006) i) the sentence has been judged to be a risk disclosure in the annual report if the coder is informed of any opportunity or prospect, or of any hazard, danger, harm, threat or exposure, that has already impacted upon the company or may impact upon the company in the future or of the management of any opportunity, prospect, hazard, harm, threat or exposure ii) and have been categorized and counted in six risk categories, namely financial risks, operations risks, integrity risk, information processing and technology risks, strategic risks, and empowerment risk (Table 1), by performing ex-post manual content analysis. Furthermore, we used Pearson correlation and Wilcoxon signed ranks (advocated as adequate practical methods, inter alia, by the closest study to ours of Linsley & Shrives, 2006) to empirically test the set hypotheses related to the firm's risk reporting practices. The Wilcoxon signed rank test is a nonparametric test used to compare two dependent samples when the differences between these two pairs of data are assumed to be non-normally distributed. In addition, non-parametric tests are more intuitive and simpler to calculate by hand when the sample size is small. We employed nonparametric test because the assumption required for parametric tests (e. g. Normality of the data) are not valid. It should be noted that with a small sample size, normality test can have insufficient statistical power to provide useful results. Therefore, the dataset for our study is also regarded as a small sample for the parametric tests, and consequently non-parametric tests were utilized.

Descriptive statistics
In the annual reports, a total of 986 risk disclosure sentences were recognized (Table 1). Table 1 shows the classifications and features of the sentences that clarify the risk disclosure. The three risk classifications with the highest number of disclosures are financial risk (289 disclosures), strategic risk (257 disclosures), and operation risk (248 disclosures). Additionally, the other three risk classifications are information processing and technology, empowerment, and integrity risk disclosure, containing 104, 54, and 34 disclosures, respectively. Descriptive statistics are reported in Table 2. The sample sizes for the total, financial, and nonfinancial risk disclosure sentences are 986, 289, and 697, respectively. Table 2 shows that the mean total risk disclosure rate per report is 29 sentences. The variation in the number of total risk disclosures is relatively large, which is in the same line with Linsley & Shrives (2006) study, ranging from a minimum of 6 sentences to a maximum of 183 sentences. In our sample, a Chinese company's total risk disclosure had the lowest rate (6 sentences), and a U.S. company's total risk disclosure had the highest rate (183 sentences). In addition, the mean value for the financial disclosure rate per report (about nine sentences) is lower than the non-financial disclosure rate per report (about 21 sentences). The range of non-financial disclosures is also relatively wide, from a low of 1 sentence in a U.S. company to a high of 164 sentences in a U.S. company. However, the variation in the number of financial risk disclosures is relatively small, ranging from the lowest financial risk disclosure rate per a report in Japan (1 sentence) to the highest financial risk disclosure rate per a report in Japan (31 sentences).

Hypothesis 1 Examination
To test the first set of hypotheses, the first three rows of Table 3 contain Pearson correlations between the first main variables of interest, especially the level of relationship between the number of risk disclosures and the independent variable of size (company size). The correlation coefficients for the measures of size, the natural log of market value, and the natural log of turnover, indicate that no significant association exists between company size and the number of total risk disclosures, the number of financial risk disclosures, and the number of non-financial risk disclosures.
Tab . Therefore, we maintain that larger companies do not disclose more risk information to manage stakeholders' perceptions about corporate reputation, as is suggested by legitimacy theory and resources-based perspective adopted in developing the first hypothesis.

Hypothesis 2 Examination
To test the second set of hypotheses, the last two rows of Table 3 comprise Pearson correlations between the second main variables of interest, especially the relationship between the number of risk disclosures and the independent variable of the level of risk. The correlation coefficients for the beta factor show no significant association between risk measures, the number of total risk disclosures, and the number of non-financial risk disclosures. The beta factor is negatively correlated with the number of financial risk disclosures (r = -0.332; p < 0.10). The negative correlation between the beta factor and the number of financial risk disclosures implies that companies with higher levels of systematic risk are closing greater volumes of financial risk information. The correlation coefficients for stock volatility are significantly correlated with the number of total (r = -0.311; p < 0.10) and non-financial (r = -0.318; p < 0.10) risk disclosures, except for financial risk disclosures. The negative correlation between the stock volatility and the number of total and non-financial risk disclosures indicates that companies with higher levels of risk are closing greater volumes of total and non-financial risk information.

Wilcoxon Signed Ranks Test
To examine Hypotheses 3-5, Wilcoxon signed rank tests were employed ( Regarding the second hypothesis, we determined that the relationship between company risk and risk disclosure is negatively associated, and consequently we do not support the idea that riskier companies disclose more risk information as suggested by agency theory as well as prior literature (Elshandidy et al., 2013;Taylor et al., 2010). Against this background, ours is the first study that confirms the negative association between the level of company risk and risk disclosure, and thus our results go against the conventional wisdom that non-financial companies with higher levels of risk disclose greater volumes of risk information. This result was robust and confirmed using both beta and stock volatility, factors which successfully addresses the issues related to risk disclosure. Both risk measures indicate the same negative sign.
A number of significant implications emerge from this evidence. Companies with higher systematic risk in the automotive industry are prone to disclose a greater amount of financial risk information (Table 3). This could be explained by the idea that companies with a higher level of risk have an incentive not to disclose financial risk information, as they may be subsequently required to justify their prior estimates. This may leave them vulnerable to litigation and, thus, a concealing motive prevails.
Additionally, the higher stock volatility of companies in the automotive industry can lead them to disclose more total and non-financial risk information. This may also lend support to the impression management argument or legitimacy theory since the results of the current study reveal that the riskier companies for stockholders disclose less, implying management potentially manipulates risk reporting by the omission of information (less complex reporting) to affect the risk perceptions of investors.

Examination of Hypothesis 3
In testing Hypothesis 3, the number of positive ranks is 20, with the number of negative ranks 14 (Table 4) . These studies indicate that monetary risk disclosure will be considerably less than the amount of non-monetary risk disclosure. However, these results were not statistically significant.

Examination of Hypothesis 4
The results of the Wilcoxon test do not support Hypothesis 4. Here the opposite is the case, as the number of future risk disclosures is significantly greater than the amount of past risk disclosure (Table 4). It can be seen that the number of negative ranks is 33, and the number of positive ranks is 1. The significance value of 0.000 implies strong support for this reverse case.
Regarding the fourth hypothesis, the findings reveal a reverse relation: future risk disclosures are significantly higher than those of past risk disclosures. There are possible explanations for this result; one of the key fundamental explanations stems from the idea that shareholders and stakeholders in the automotive industry may require companies to disclose information concerning their future predictions, as the economic environment is considered dynamic and thus they cannot rely on past information (Kılıç & Kuzey;.

Examination of Hypothesis 5
The Wilcoxon test results do not support Hypothesis 5, i.e. that there would not be a significant difference between good and bad risk disclosures. There is strong support for the difference between bad and good risk disclosures (p=0.015), by which 23 companies disclosed a higher number of bad risk disclosures, compared to 11 companies that disclosed more good risk disclosures (Table 4).
We again found reverse results regarding the fifth hypothesis which show a significant difference between bad and good risk disclosure in the automotive industry. This idea stems from the attribution theory, i.e. that directors in the automotive industry convey bad news to third parties, particularly when discussing bad risk news. Table 5 shows a summary of the previously described outcomes of hypothesis testing (see Appendix A).

CONCLUSION
Prior research has declared that increased corporate transparency facilitates firm competitiveness (Wyld, 2009;Shivaani & Agarwal, 2020). In other words, firm-specific information revealed to the public helps make enterprises more competitive. Wyld (2009) indicates that less transparent firms have lower growth rates, lower investment efficiency, and higher capital costs. Furthermore, less transparent firms provide less informed choices for shareholders. Enhanced transparency alerts stakeholders that a business aiming to acquire a competitive position continues to maintain its ethical standards (Shivaani & Agarwal, 2020). Despite a substantial increase in risk reporting studies in recent years as well as significant interest in a wide range of user groups, the corporate transparency phenomenon in terms of risk disclosures is still very much in its infancy (Abraham & Shrives, 2014).
This study examines the level and context of risk disclosure of publicly traded automotive companies by investigating the association between the level of risk disclosure and firm-specific charac-teristics (company size and company risk), as well as exploring the risk disclosure characteristics (namely, type of risk, time orientation, and proportion of good and bad risk disclosure). In its examinations of the risk reporting practices of non-financial companies using textual analysis, our work seeks to extend in several ways research in mainstream financial accounting. First, our cross-country, single-industry (automotive) study addresses the lack in the literature regarding industry-specific risk, cross-country, and automotive industry-based disclosure studies. Second, our findings in terms of the relationship between the level of risk disclosure and company level of size/risk can help develop further hypotheses which can shed some more light on the question of whether larger and more volatile listed companies reveal more risk information. Finally, our explanation of the prevalent disclosure context can serve policymakers, academicians, and users of annual reports to reach a deeper understanding of the phenomenon of current risk reporting practices of non-financial companies, as well as the limits and usefulness of risk disclosure.
This study presents some limitations that could be overcome by further studies. We here focus on annual reports; for deeper and more enhanced research it would be useful to gather and evaluate other evidence in supplementary documents such as corporate press releases, websites, prospectuses and presentations. Additionally, the Wilcoxon signed rank test that we employed in our study has a noticeable disadvantage in that it disregards the magnitude of the observations measured. While it distributes a sign to each observation to indicate whether it remains the below or above hypothesized value, this does not account for scale or significance. The statistical power of the Wilcoxon signed rank test is thus limited, and further studies should focus on more parametric methods. Another limit of our research which might be addressed in future investigations is that our work here does not represent a longitudinal study. Further studies could apply automated content analysis as well as explore the risk reporting practices of listed non-financial companies in a longitudinal framework using a greater number of proxies to focus on information and agency problems. Moreover, studies might evaluate the effects of risk reporting practices on user risk perceptions of financial information. Other research projects might attempt to explore the motivations for both voluntary and mandatory risk disclosure in non-financial listed companies. Each of these potential tracks could in their own way contribute to a more complete understanding of information content and risk factor disclosure.